I've used freesurfer to segment (recon-all)  approximately 30 different subjects from a study. I've noticed that a large number of these subjects have a misclassification of the white and pial boundaries in the occipital lobe (such that grey matter is being classified as white matter). I believe this problem is originating from my scanner and original data. I'm acquiring a T1W BRAVO sequence on a GE Discovery 750w 3T scanner, using a 32 channel phased array head coil. For many subjects, this acquistion gives intensity inhomogeneity in the occipital lobe.  I've begun using a GE based correction technique (PURE) on the images, and it appears to improve the intensity in the occipital lobe, but Freesurfer is still frequently misclassifying. 

I'm wondering what would be the best way to edit the data. I've often tried to edit the wm.vol, but this doesn't seem to fix anything. For the most part, the wm.vol is correct to begin with, but the wm cutting line isn't segmenting with the wm.vol. I've also tried to add control points but this doesn't seem to improve anything either.

I've attached an image which I hope will further explain my problems. 

Thanks for any help